Abstract A031: Cross-platform integration of serum NMR and MRI radiomics identifies robust biomarkers for kidney cancer

Abstract Background: Kidney cancer comprises biologically diverse malignancies—including clear cell renal cell carcinoma (ccRCC), papillary RCC (pRCC), and chromophobe RCC (chRCC)—each demonstrating distinct metabolic dependencies and radiologic phenotypes. Current diagnostic approaches rely heavily on imaging and histopathology, with limited ability to detect early-stage disease or predict progression. Integrating serum-based metabolomics with MRI-derived radiomic signatures offers a promising non-invasive strategy to identify stage-specific biomarkers reflecting tumor metabolism, microenvironmental remodeling, and anatomical heterogeneity. Methods: Serum samples were collected from patients across major RCC subtypes (n = 256) at stages I–IV prior to treatment. Proton nuclear magnetic resonance (^1H-NMR) spectroscopy was performed using a standardized NOESY-presat protocol. Metabolites were quantified via Chenomx/Bruker profiling and normalized using PQN. In parallel, multiparametric MRI (T1, T2, DWI, and contrast-enhanced sequences) was processed using PyRadiomics to extract shape, intensity, and texture features. Multivariate modeling (OPLS-DA, LASSO, and Random Forest) was used to identify discriminant metabolic and radiomic signatures. Integrative analysis employed canonical correlation analysis and multi-omics factor analysis (MOFA). Results: Distinct metabolic fingerprints differentiated RCC subtypes and stages. Early-stage disease showed elevation of citrate, creatine, and branched-chain amino acids, whereas late-stage and metastatic RCC displayed increased lactate, glutamine, kynurenine, and lipid-derived metabolites, indicating a shift toward aerobic glycolysis and immune-metabolic reprogramming. MRI radiomics identified stage-dependent textural heterogeneity (GLCM entropy, GLSZM zone variance) strongly associated with metabolic dysregulation (r = 0.61–0.78). Integrative modeling yielded a robust 12-feature biomarker panel combining 7 metabolites and 5 radiomic descriptors that discriminated early vs. advanced RCC with an AUC of 0.92 and subtype classification accuracy of 87%. Preliminary pathway enrichment implicated deregulated TCA cycle, glutaminolysis, and tryptophan metabolism. Conclusions: This study provides a comprehensive integrative metabolomic–radiomic framework that identifies non-invasive biomarkers for RCC detection, subtype classification, and stage prediction. The combined NMR–MRI signature has strong translational potential for personalized risk stratification, surveillance, and therapeutic decision-making in kidney cancer. Citation Format: poornima verma. Cross-platform integration of serum NMR and MRI radiomics identifies robust biomarkers for kidney cancer [abstract]. In: Proceedings of the AACR Special Conference in Cancer Research: Innovations in Kidney Cancer Research: From Molecular Insights to Therapeutic Breakthroughs; 2026 Mar 13-16; Philadelphia, PA. Philadelphia (PA): AACR; Cancer Res 2026;86(5_Suppl_2):Abstract nr A031.